Table 1: Reliability and factor
analysis results.
|
No
|
Component
|
Number
of items
|
Cronbach’s
Alpha
|
Total
variance explained
|
Conclusion
|
|
1
|
Reliability (DTC)
|
5
|
0,847
|
62,523%
|
Accepted
|
|
2
|
Availability responsiveness (DU)
|
6
|
0,869
|
60,706%
|
|
3
|
Assurance (DB)
|
5
|
0,909
|
73,329%
|
|
4
|
Empathy (TC)
|
4
|
0,926
|
81,964%
|
|
5
|
Tangibles (HH)
|
5
|
0,896
|
70,969%
|
|
6
|
Health Consciousness (SK)
|
5
|
0,844
|
62,983
|
|
7
|
Attitude (TD)
|
6
|
0,850
|
58,319%
|
|
8
|
Intention (YD)
|
4
|
0,861
|
70,913%
|
Table 1 provided Cronbach’s Alpha values and exploratory factor analysis
(EFA) results. The Cronbach’s Alpha results for all constructs were above the
recommended level of 0.6 (Hair & ctg), and total variance explained results
were above 50%. Simultaneously, the authors created average values representing
quality perception factors, using for analysing exploratory factor EFA and
valid SEM model in the following steps. The group of factors needed to create
new identity values was as follows: Reliability (DTC), Availability
Responsiveness (DU), Assurance (DB), Sympathy (TC), Tangible Factor (HH).
Exploratory
Factor Analysis (EFA) results
EFA results stopped
at the third time with KMO statistics at 0.878 and Bartlett test value with
Sig. = 0.000 (<0.005). This demonstrated that observed variables generally
correlate. Specifically, in the first attempt, Factor Loading of TD1, TD2, SK5
did not appear, therefore, the authors eliminated these variables from the
model. In the next attempt, Factor Loading of TD3 did not appear so they were
eliminated from the model either. After eliminating those unqualified
variables, there were four groups of factors left, including 16 observed
variables. The Eigenvalue was 1.129, and the Total Cumulative Variance
explained was 67.677% (Above 50%), which meant at the value of Eigenvalue of
1,129, these 4 components could explain 67,677% of variance of data collected
(Table 2).
Table 2: EFA results.
|
Component
|
Items
|
|
Perception of quality
(NTCL)
|
DTC, DU, DB, TC, HH
|
|
Health consciousness
(SK)
|
SK1, SK2, SK3, SK4
|
|
Attitude (TD)
|
TD4, TD5, TD6
|
|
Intention (YD)
|
YD1, YD2, YD3, YD4
|
Table 3: Measurement Model
analyzing results.
|
Component
|
Cronbach’s Alpha
|
Composite Reliability
|
Average variance
extracted (AVE)
|
|
NTCL
|
0.906
|
0.930
|
0.727
|
|
SK
|
0.848
|
0.898
|
0.687
|
|
TD
|
0.866
|
0.918
|
0.788
|
|
YD
|
0.863
|
0.906
|
0.708
|
Structural
Equation Modelling (SEM) analysis
Measurement model testing (Table 3): In order to test the scale’s
reliability, the research use Composite Reliability (CR), Average variance
extracted (AVE), Outer loading. The Composite Reliability and outer loading
must be above 0.7 and 0.4 respectively [15]. Besides, according to, the average
variance extracted must be above 0.5 to prove the reliability and Convergent
validity [16]. The minimum value is 0.687 from the Health Consciousness. The
results indicated satisfactory of all the values in measurement model.
Structural model
testing: SEM results in image 4 illustrated that model
had Chi-Square Test at 1072,1493 with p-value = 0.000 < 0.005. However,
according to, if the model has the SRMR value <0.1, then it will be
considered appropriate for the real data. Hence, SRMR = 0.079 < 0.1, the
model was considered appropriate with the data collected in Hanoi [17] (Figure
2).
Figure
2:
PLS-SEM Results analysis.
According to the result, perception of quality and health consciousness
only explained 35,9% of attitude variation, at the same time, perception of
quality, health consciousness, and attitude explain 55.9% of intention variation,
at the meaning level of 5%. However, according to Chin and associates (1996),
when analysing the effect of independent variables towards dependent variable,
the researcher not only consider the relationships of meaning level but also analyse the effect level. Therefore,
this study continued to test bootstrapping [18].
Bootstrapping
testing
From 5000 observations, all the original weights
are in the range of 95%. Therefore, the estimates in the model were considered
to be reliable (Table 4)
Table 4: Bootstrapping
structural model testing.
|
|
Original
Sample (O)
|
Sample Mean (M)
|
Standard Deviation
(STDEV)
|
5%
|
95%
|
|
NTCL -> TD
|
0.363
|
0.363
|
0.000
|
0.277
|
0.448
|
|
NTCL -> YD
|
0.570
|
0.568
|
-0.002
|
0.503
|
0.636
|
|
SK -> TD
|
0.327
|
0.329
|
0.002
|
0.232
|
0.406
|
|
SK -> YD
|
0.132
|
0.133
|
0.001
|
0.050
|
0.209
|
|
TD -> YD
|
0.469
|
0.470
|
0.000
|
0.384
|
0.553
|
Table 5: Hypothesis testing
results.
|
|
Original
Sample
|
T-value
|
P-value
|
Hypothesis
|
|
NTCL -> TD
|
0.363
|
6.941
|
0.000
|
Accepted H1
|
|
NTCL -> YD
|
0.568
|
14.122
|
0.000
|
Accepted H2
|
|
SK -> TD
|
0.329
|
6.264
|
0.000
|
Accepted H3
|
|
SK -> YD
|
0.133
|
2.715
|
0.003
|
Accepted H4
|
|
TD -> YD
|
0.470
|
9.043
|
0.000
|
Accepted H5
|
Hypothesis
testing (Table 5)
Regression
analysis for hypothesis
In this research, the authors used the standardized regression equation
to test the effect of independent factors on dependent ones because the standardized regression equation
has the economic meaning more rather than mathematical meaning (Table 6).
According to the results, the three factors all
had influences on behavioral intention, specifically, perception of quality was
the most important factor. Besides, p-value was 0.000, indicating that the
estimated data was suitable to real data. R-Square coefficient was 0.513,
showing that these 3 factors explained 51.3% of the variance of use intentions.
The dustbin-Watson = 1,763 and VIF coefficient of the three factors are all
below 5. Therefore, we have the regression equation: YD = 0,532NTCL+0,192TD
+0,102SK. Hence, the importance of each factor on behavioral intention
respectively were: perception of quality, attitude and health consciousness.